[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: io memory mapping on alphas



	Don't mind me.  I was in the process of reading IO-mapping.txt
when I wrote this.  I got to the section that answered it just _after_ I
sent the message. <sigh> Teach me to ask too soon.

Advice on fixing it?  I don't think this is going to be fun...

- Bob

On Sun, 8 Nov 1998, Bibek Sahu wrote:

> 
> 	I can't remember if my message got through or not, but I haven't
> seen it.  In either case, I'm still searching for answers.
> 
> 	What does it take to access memory-mapped io on hardware?  Where
> can I find some details on it?  (the kernel docs haven't been that great,
> nor have the header files I've looked through so far; haven't gotten to
> the C files yet...)
> 
> 	The problem is this:
> 	I have a Matrox Millennium II video card.  It has memory mapped
> stuff at 0xa000000, 0xb000000, and 0xb800000.  The first is the
> framebuffer.
> 
> 	The cvs tree at vger likes to dereference NULL pointers as does
> 2.1.127, so I'm currently using 2.1.126 on my sx164.  In the vger tree is
> a matroxfb (kernel framebuffer module) which I've backported to 2.1.126.
> 
> 	I've also been working with GGI, trying to get kgicon (kgi for the
> kernel framebuffer) working.
> 
> 	Both of these successfully find the card.  However, problems arise
> when matroxfb tries to access memory at 0xb001fde and when kgicon tries to
> access memory at 0xa000000.  matroxfb is trying to get something out of
> the second memory mapped range (which I think contains registers), kgicon
> is trying to access the framebuffer.
> 
> 	In both cases, the kernel Oopses with
> 
> Nov  8 21:27:37 prophet kernel: Unable to handle kernel paging request at
> virtual address 000000000b001fde
> 
> 	(or a000000, as appropriate).
> 
> 	What I want to know is: why is the kernel trying to page the IO
> space?  Is it supposed to be doing that?
> 
> 	I admittedly don't know much about memory, which is why I'm
> asking, but that seems to me to be terribly wrong.  This presumably
> doesn't happen on x86 machines; I'd imagine these modules load on at least
> _somebody's_ machine (and I know that at least kgicon does). 
> 
> 	If you don't have an answer, but can point me in a direction where
> I might find some answers and/or memory mapping info, I'd appreciate that,
> too.
> 
> System:
> Microway Screamer SX, 21164PC
> 64MB ram
> 4GB IDE Hard Drive
> Kernel 2.1.126 + international crypto patch + sx IDE controller dma patch
> 
> 	As I said, 2.1.127 doesn't work on my machine.  I haven't tried
> downgrading to older kernels yet.
> 
> Thanks for any and all help!
> - Bob
> 
> -- 
> To unsubscribe: send e-mail to axp-list-request@redhat.com with
> 'unsubscribe' as the subject.  Do not send it to axp-list@redhat.com
> 



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index] []